Practical Microservices Jonathan Schabowsky, Sr. Architect, Office of the CTO

What are Microservices? Small in Size, Single in Purpose Communicate using technology-agnostic protocols Independently Deployable Released via Automated Processes

Microservice Architecture Desires Scalable: Horizontal – Services must scale in order to meet demand Dynamic Scaling – Change number of running instances based on actual demand Shock Absorption – Properly deal with irregular bursts of utilization Fault tolerant: In case of failure or exception unit of work should be retried Resilient to unexpected IaaS/DC outages Performant: User experience dictates that applications do not take a long time to return Service execution must not exceed planned processing/response time Manageable: Visibility into application performance and/or bottlenecks Secure: Communications between Services need to be encrypted when using Public Clouds

The Event: Dalmore King Alexander III is Amazing! Wonder if I can buy it in Utah?!

The Problem: Department of Alcohol Control AKA: The State Liquor store WebApps: OnlinePriceList InventoryQuery

OnlinePriceList Data (16,172 Products):

InventoryQuery Data:

The Outcome: Department of Alcohol Control AKA: The State Liquor store WebApps: OnlinePriceList InventoryQuery Wouldn’t it be great if I could: see when it was last in-stock be notified when in-stock

On Premise Environment On Premise Environment amazon web services Elastic Stack Utah Department of Alcohol Control OnlinePriceList InventoryQuery Deployment problem On Premise Environment On Premise Environment

Application Iteration#1: Department of Alcohol Control AKA: The State Liquor store OnlinePriceList InventoryQuery WebApps: AWS Data Lake Elastic Stack http POST http POST UtahDABC-Client SpringRestTemplate Business Logic UtahDABC-Inventory SpringRestTemplate Business Logic Spring RestServices http REST POST http REST POST

Application Iteration#1 Issues: Department of Alcohol Control AKA: The State Liquor store OnlinePriceList InventoryQuery WebApps: Tight Coupling: What if I want those events to also go to ELK? AWS Data Lake Service Discovery: How do I know where ELK is? Elastic Stack The Client’s Processing time is dependent on all downstream services! http POST http POST Tight Coupling: What if Inventory Service is Slow? Or down? UtahDABC-Client SpringRestTemplate Business Logic UtahDABC-Inventory SpringRestTemplate Business Logic Spring RestServices http REST POST http REST POST Performing Query for non Whiskey Products!

Not everything is a Nail When holding Hammer… REST is a Hammer Not everything is a Nail Good for: Synchronous Interactions Externally facing APIs Baggage it brings: Blocking Service Coupling Complex Failure Scenarios – Responsibility

Think Event Driven Microservices have inputs and outputs Inputs and outputs can be “events” So what is an Event? an event is an action or occurrence recognized by software, often originating asynchronously from the external environment, that may be handled by the software Stop thinking about services INVOKING services! Instead our services emit events! What/where do we emit these events to? Messaging (preferability Solace!)

Why Solace? Technical Characteristics http://dev.solace.com/ Open APIs (AMQP1.0, MQTT, REST, JMS) Native APIs (JMS, Java, .NET, Node.js, JS, C#) Extensive Spring Support Native integration to all Clouds, PaaS, IaaS WAN routing & optimization Broad 3rd party app integrations Performance 80K guaranteed msgs/sec P2P >600K guaranteed msgs/sec fanout Scalability 100K IoT connections – can scale to billions Robustness Security Rich Management

Messaging Terminology Topic In Solace, a Topic is a property of a message.” That’s it! It’s just a text string published in the message header It’s completely dynamic Supports hierarchies using the ‘/’ delimiter Supports wildcard subscriptions ‘*” replaces any part or whole level identifier “>” replaces any number of level identifiers Queue Can have topic subscriptions to “attract” messages Exclusivity – Like Dating Exclusive – Only one connection gets all messages Non-Exclusive – Round Robin messages across ALL connections Durability – Will hold messages even if there are no active consumers Enforce TTL – can expire messages

OnlinePriceList Data (16,172 Products): Topic Hierarchy!

Spring MessageListenerContainer The Better Way Demo: Department of Alcohol Control AKA: The State Liquor store OnlinePriceList InventoryQuery WebApps: AWS Data Lake Elastic Stack http REST POST http POST PCF Tile RDP UtahDABC-Inventory SpringJmsTemplate Business Logic Spring MessageListenerContainer UtahDABC-Client http POST Queues: -productQforInventory Topic Sub: product/A/W/> -toELK-Inventory Topic Sub: inventory/> Business Logic JMS – topic: product/X/Y/Z JMS – topic: inventory/X/Y/Z SpringJmsTemplate

Data Flow: Solutions: Elastic Stack Utah Department of Alcohol Control amazon web services Data Flow: Elastic Stack http REST POST JMS JMS Utah Department of Alcohol Control Solutions: Legacy-to-cloud Cloud-to-Cloud PaaS Support Cloud network outbound cost? OnlinePriceList InventoryQuery Deployment problem On Premise Environment On Premise Environment

Utilizing Compression: amazon web services Utilizing Compression: Elastic Stack http REST POST JMS Compression <Bridge> JMS Utah Department of Alcohol Control Reduces Network Cost! (80%+ Savings on Bandwidth Utilization) OnlinePriceList InventoryQuery Deployment problem On Premise Environment On Premise Environment

Event Driven Interaction Considerations Benefits Limited service to service coupling (excludes data level) No impact service upgrades Easy to create new services and fail fast Resilient to service failures Generally more resource efficient Enables fine grained service scaling Challenges: Not as Natural – Many architects and developers think in a “synchronous” way. Design pattern education is required. Correlation – Services must be able to correlate different events since they occur asynchronously